Next: Implementation
Up: Design
Previous: Request/Response handling
The Cache Management uses the same netpipes as were mentioned in the
previous section. Instead of spooling the data to a socket, they
simply spool it directly to disk in the same manner, using file descriptors.
I decided to have a single level cache with random filenames mainly
for two reasons:
- 1.
- This will make it harder for a network operator to snoop on the
client, i.e. he'll have a hard time deciphering what each client has
been viewing by just browsing through the cache.
Ideally, the cache would be encrypted, but that will have to wait for
a future project.
- 2.
- By storing the files in a single directory some time will be
saved in fetching the files from the cache, compared to storing the
files in a hierarchical manner.
The cache will be stored in a hash-table, which stores keys and
filenames as a name/value pair.
Magnus Bjornsson
1998-05-08